MediaTeam Speech Corpus: a first large Finnish emotional speech database

In this paper, a large Finnish emotional speech database is introduced. The database contains simulated emotional speech, reflecting some of the affective states commonly known as basic emotions. The database has been analyzed instrumentally in terms of some 40 acoustic/prosodic parameters, and is currently being annotated in terms of a number of linguistically relevant contrasts. The annotation involves several hierarchical tiers, relating, for example, to the distribution of sentence accents and different types of pauses in the data. Furthermore, listening tests at several levels of abstraction are being carried out to verify that the speech samples contain the intended emotions. Once the analysis is complete, the results can be utilized in content-based information retrieval from audio databases,